Solving the Feldstein-Horioka Puzzle With Financial Frictions

Long-run average savings and investment rates are strongly correlated across countries, which Feldstein and Horioka (1980) read as evidence of substantial frictions in international capital markets. This paper does two things with that reading. First, it verifies the conjecture on which it rests, showing in a calibrated stochastic general equilibrium model of a continuum of small open economies that frictionless complete markets do imply a cross-country savings-investment coefficient of essentially zero (−0.01) – a point the authors say “is new to the literature,” since prior theoretical work had mostly addressed the time-series correlation, which arises with or without frictions. Second, it asks which financial frictions can produce the observed coefficient and the observed volume of capital flows at the same time. Two are considered: limited enforcement, where contracts are backed only by the threat of permanent exclusion from financial markets plus an output loss, and limited spanning, where the only tradable asset is a noncontingent bond. Neither works alone. The frictionless benchmark generates an average absolute current-account-to-GDP ratio of 62 percent, roughly nine times the 7 percent in the data, and a foreign-asset position of 6.12 times GDP against 0.49. Limited enforcement with a full set of contingent assets barely restricts anything, because continued market access is so valuable under volatile productivity shocks that default is unattractive: the coefficient stays at −0.01 and capital flows at 56 percent, and even setting the output loss to zero and letting defaulters re-enter markets with certainty only lifts the coefficient to 0.23 while collapsing flows well below the data. Limited spanning with natural debt limits gives a coefficient of 0.05 and flows of 38 percent; tightening the limits exogenously to a fraction of resources does reproduce the data, as Castro (2005) had shown, but leaves the source of the limits unexplained. Combined, the two frictions generate endogenous noncontingent debt limits that are tight for two distinct reasons – they must hold under the worst realization of next period’s shock, and the value of staying in markets is lower when only a bond can be traded, so default is more tempting – permitting borrowing of only about 30 percent of output. The calibrated two-friction model then produces a Feldstein-Horioka coefficient of 0.52 with a standard error of 0.05 against 0.52 in the data, a capital flow ratio of 10 percent against 7 percent, a foreign-asset ratio of 0.40 against 0.49, a savings-investment correlation of 0.77 exactly as observed, cross-country dispersions of savings and investment rates of 0.06 and 0.04, and a degree of international risk sharing far closer to the data than the other models deliver. Two mechanisms are identified as essential. The tight limits force low-capital countries to save in order to invest and, through the interest rate, discourage high-capital countries from lending, generating the positive cross-country correlation. And because limited spanning makes repayment obligations noncontingent, enforcement constraints bind in bad states rather than good ones, which eliminates the enforcement model’s counterfactual implication that investment can rise when a country is hit by a bad shock. The authors are explicit about what the results depend on: the quantitative conclusions are sensitive to default penalties – raising the output loss from zero to 2 percent moves the coefficient from 0.94 to 0.45 – and they assume throughout that defaulters have their debt fully written off and are treated on re-entry like countries that never defaulted, which if relaxed would loosen borrowing limits and lower the coefficient.

Jing Zhang | Macro Paper Warehouse

Solving the Feldstein-Horioka Puzzle With Financial Frictions